David Levine has an interesting article on neuroeconomics (entitled Neuroeconomics?). While skimming through the article I found an interesting analogy concerning neuroeconomics.I did not yet read the article (will do that soon). Nevertheless, here are some remarks concerning the analogy.
David Levine argues:
“Suppose you wanted to study Microsoft Word in order, say, to build a better word processor. Would you study the CPU of a PC? Would you study how the RAM is wired? Or the ASICS? Would you study the binary code? Surely you would do none of those things. Would you study the source code? Probably not even that, most likely you would use the program, observe how it worked, and figure out how to build a program that did the same thing. That is what economists do. We observe human behavior and figure out models that behave the “way humans do.” We have no reason to believe that better understanding the wiring of the brain would improve our models any more than understanding the micro-code on an x86 chip would improve lead to improvements in word processors.” (Source)
I am not a computer scientist, yet this sounds like a bad analogy.
Concerning software (e.g., a word processor) the answer to Levine’s question is, it depends. First you need to ask whether the particular software you have is dependent on the platform it operates, or not. There are certainly a lot of platform independent software. But some software will not work if you move it to another computer — say from a Linux based system to a Windows based system, or from a 64bit system to a 32bit system. Sometimes, in order to understand how a software works and how you may improve it, you need to understand the operating system and the architecture of the processor. But this is not where the analogy fails.
Analogy fails, because it bypasses at least two arguments for neuroeconomics. Firstly, neuroeconomics shows that economic behavior may be context-dependent. Hence, explaining particular economic phenomena may require an understanding how context influences economic behavior. And in order to develop a good model of context dependent economic behavior it may be useful to understand how brain “produces” economic behavior under different conditions. If your word processor behaves differently when it is simultaneously running a Matlab code in the background, it may be good idea to inquire and understand how Matlab and Word interacts at a lower level (e.g., at the level of the operating system, chipset, processor). Secondly, even if neural level correlates of economic behavior is well in line with the assumptions of economic theory, a better understanding of how brain works in relation to economic behavior may improve our understanding of economic phenomena. Here is the analogy for this. Let us assume that there are two coders John and Mike. John only knows a couple of platform independent programing languages. He has no understanding of operating systems and computer architecture. Nevertheless, thanks to platform independent programing languages, he produces functional software for the Java platform or for Adobe Air. Mike, on the other hand, knows how a computer works, how platform independent programing languages interact with the operating system and how processors react to commands coming from the operating system. Mike, too, writes functional software for Java and Adobe Air. In terms of what they do John and Mike are equivalent. However, in terms of their understanding of what they do they are not. Mike knows better. His understanding of how software works is superior to that John. Economists who ignore neuroeconomics is like John. They produce functional models of economic behavior. However, if they would like to have a better understanding of economic behavior, they are well advised to learn more about how lower levels work by reading (and not ignoring)neuroeconomics. Remember, Levine argues:
“We have no reason to believe that better understanding the wiring of the brain would improve our models any more than understanding the micro-code on an x86 chip would improve lead to improvements in word processors.”
Well, this is partly true. If we know that our economic models are correct, understanding the neural-level will not improve our models. But even if our models are correct, a better understanding of neural correlates of economic behavior will improve our understanding of economic phenomena. Also note that we do not know whether all of our economic models are correct or not. Thus, neuroeconomics may also help us develop better models of economic behavior.
As I have said, I am not a computer scientist. So, probably I have made many mistakes in my software-hardware analogies. My computer-analogy free argument concerning how neuroeconomics may be useful for economists is here: Neuroeconomics: more than inspiration, less than revolution, Journal of Economic Methodology, 2010, 17(2): 159-169.